Hybrid Approach for the Interpretation of Nominal Compounds using Ontology

18913

Authors

nan

Abstract

Understanding and interpretation of nominal compounds has been a long-standing area of interest in NLP research for various rea- sons. (1) Nominal compounds occur fre- quently in most languages. (2) Compound- ing is an extremely productive word formation phenomenon. (3) Compounds contain implicit semantic relations between their constituent nouns. Most approaches that have been pro- posed so far concentrate on building statisti- cal models using machine learning techniques and rely on large-scale, domain-specific or open- domain knowledge bases. In this paper we present a novel approach that combines the use of lexical hierarchies such as PurposeNet and WordNet, with WordNet-based similar- ity measures for the interpretation of domain- specific nominal compounds. We aim at build- ing a robust system that can handle most of the commonly occurring English bigram nominal compounds within the domain. 1

Keywords

nan

Full Content

Copyright 2012 by Sruti Rallapalli and Soma Paul 26th Pacific Asia Conference on Language,Information and Computation pages 554–563 A Hybrid Approach for the Interpretation of Nominal Compounds using Ontology Sruti Rallapalli, Soma Paul Language Technologies Research Center International Institute of Information Technology Hyderabad sruti@students.iiit.ac.in soma@iiit.ac.in Abstract Understanding and interpretation of nominal compounds has been a long-standing area of interest in NLP research for various rea- sons. (1) Nominal compounds occur fre- quently in most languages. (2) Compound- ing is an extremely productive word formation phenomenon. (3) Compounds contain implicit semantic relations between their constituent nouns. Most approaches that have been pro- posed so far concentrate on building statisti- cal models using machine learning techniques and rely on large-scale, domain-specific or open- domain knowledge bases. In this paper we present a novel approach that combines the use of lexical hierarchies such as PurposeNet and WordNet, with WordNet-based similar- ity measures for the interpretation of domain- specific nominal compounds. We aim at build- ing a robust system that can handle most of the commonly occurring English bigram nominal compounds within the domain. 1 Introduction Understanding and interpretation of nominal com- pounds has been a long-standing area of interest in NLP research. The main reasons that make under- standing compound nouns an interesting and chal- lenging task are: (1) Compound nouns are a frequent phenomenon in many languages, occurring in differ- ent languages with varying frequencies. English and Sanskrit are two languages that display great flexi- bility in compounding ( ´O S´eaghdha, 2008). About 3.9 % of the words in Reuters are bigram nominal compounds (Baldwin and Tanaka, 2004). (2) Com- pounding is a recursive process that can lead to for- mation of large and complex compounds, that are difficult for comprehension. (3) Compounds usu- ally carry an implicit meaning that may sometimes differ significantly from that of the combining con- cepts. Consider the example of a garden knife. A garden knife is interpreted as a knife used in the gar- den. Here the modifier garden modifies the locative information of the head noun knife. Alternatively, consider the example of a gamma knife. A gamma knife is a device used to treat brain tumors by admin- istering gamma radiations in a particular manner. Here, the modifier does not necessarily modify the head, instead they both combine together to denote a different concept. This understanding of the dif- ference in the structure and purpose between gamma knife and other kinds of knife cannot be achieved by any means of statistical predictions or morphologi- cal and syntactic analyses of the compound. The most common representations adopted for the interpretation of nominal compounds involve an in- ventory of verbs, prepositions or abstract semantic relations. Verb and preposition paraphrases bring in lexical ambiguity in the interpretation of the com- pound, essentially owing to the polysemous nature of verbs and prepositions. For example, morning tea and bar lights would both be paraphrased using the preposition in. However, the paraphrase ’tea in the morning’ conveys the temporal aspect of …

Content truncated to 500 words

Labeled Text

Understanding and interpretation of nominal compounds has been a long-standing area of interest in NLP research for various reasons. (1) Nominal compounds occur frequently in most languages. (2) Compounding is an extremely productive word formation phenomenon. (3) Compounds contain implicit semantic relations between their constituent nouns. Most approaches that have been proposed so far concentrate on building statistical models using machine learning techniques and rely on large-scale, domain-specific or open- domain knowledge bases. In this paper we present a novel approach that combines the use of lexical hierarchies such as PurposeNet and WordNet, with WordNet-based similarity measures for the interpretation of domainspecific nominal compounds. We aim at building a robust system that can handle most of the commonly occurring English bigram nominal compounds within the domain. 1

Publication Details

Paper ID:
18913
Year:
2012
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan